The 10 Research Topics in the Internet of Things

The 10 Research Topics in the Internet of Things

arXiv:2012.01594v1 [cs.DC] 2 Dec 2020

Wei Emma Zhang1, Quan Z. Sheng2, Adnan Mahmood2, Dai Hoang Tran2, Munazza Zaib2, Salma Abdalla Hamad2, Abdulwahab Aljubairy2, Ahoud Abdulrahmn F. Alhazmi2, Subhash Sagar2, and Congbo Ma1

1School of Computer Science, The University of Adelaide, SA 5005, Australia

Email: wei.e.zhang@adelaide.edu.au 2Department of Computing, Macquarie University, NSW 2109, Australia

Email: michael.sheng@mq.edu.au

Abstract--Since the term first coined in 1999 by Kevin Ashton, the Internet of Things (IoT) has gained significant momentum as a technology to connect physical objects to the Internet and to facilitate machine-to-human and machine-to-machine communications. Over the past two decades, IoT has been an active area of research and development endeavors by many technical and commercial communities. Yet, IoT technology is still not mature and many issues need to be addressed. In this paper, we identify 10 key research topics and discuss the research problems and opportunities within these topics.

Index Terms--Internet of Things, Energy Harvesting, Recommendation, Search, Summarization, Conversational IoT, IoT Service Discovery

I. INTRODUCTION

The vision of a connected and smart world can be traced back to 1920s, as explained by Nikola Tesla in 1926: "When wireless is perfectly applied the whole earth will be converted into a huge brain, which in fact it is, all things being particles of a real and rhythmic whole. We shall be able to communicate with one another instantly, irrespective of distance. . . . A man will be able to carry one in his vest pocket" [1]. However, the term "Internet of Things" (IoT) was only first coined in 1999 by MIT's Kevin Ashton when he promoted the radio frequency identification (RFID) technology. Since then, IoT has received significant momentum as a promising technology to turn each physical object (i.e., a thing) into a node on the Internet and to facilitate machine-to-human and machine-tomachine communication with the physical world [2]?[4]. By connecting and integrating both digital and physical entities, IoT enables a whole new class of exciting applications and services such as smart cities, smart homes, Industry 4.0, and Society 5.01.

Over the past two decades, particularly the last 10 years, IoT has been a thriving area of research and development efforts, with a quickly rising body of produced research work. According to Microsoft Academic2, there were only 26 publications on IoT in 2000 and 160 publications in 2009. The publication number, however, dramatically increased over the past few years, e.g., 10,926 in 2016, 15,765 in 2017, 21,906 in 2018, and 26,885 in 2019. Despite of these exciting

1 0/index.html 2.

Fig. 1. A generic architecture of IoT systems. The physical layer collects data to be processed and modeled in the data layer, which then delivers the data to the application layer.

activities, IoT techniques still remain immature and many technical hurdles need to be overcome.

The aim of this paper is to identify several important IoT research topics and areas, ranging from energy harvesting, data analytics, search, recommendation, security, privacy and trust in IoT, as well as the topics that arise in the adoption of new computing paradigms such as social computing, service computing, edge computing, and artificial intelligence (e.g., conversational AI and text/video summarization).

II. RESEARCH DIRECTIONS IN IOT

Over the last two decades, there are a broad spectrum of activities around IoT research and development. In this paper, we highlight 10 topic areas that span across the three main layers of IoT architecture as illustrated in Figure 1. We believe these topics represent the most important research efforts from the community. It should be noted that this is not the complete list of the important research topics. Many other topics such as standard development and regulatory implications are outside the scope of our discussions.

A. Energy Harvesting

The rapid evolution in the promising paradigm of IoT has resulted in a massive distributed network of intelligent objects possessing a highly varying compute, storage, and networking

capabilities. These networked objects interact with one another primarily in a bid to exchange a diverse range of information having a direct influence for enhancing the quality of our daily lives by ensuring seamless access to smart services anywhere at anytime. However, a number of IoT sensors and embedded IoT devices have a limited lifespan since they are powered by batteries and, therefore, requires replacement periodically (e.g., every few years) making this an inefficient, laborious, and costly process. Smart energy management, especially, energy harvesting (also referred to as energy scavenging), is indispensable for ensuring energy efficiency in IoT objects [5].

Energy harvesting is a mechanism for transforming readily available energy from natural or artificial resources into usable electrical energy. It comprises four salient phases, i.e., selection of an optimum and abundantly available energy resource, its transformation, storage, and consumption. Some of the energy sources that could be harvested for IoT include, but are not limited to, thermal energy, light energy, RF energy, electromagnetic energy, chemical energy, and mechanical energy. For transformation purposes, corresponding harvesters or transducers are employed to identify and transform energy. In the case of storage, rechargeable batteries and super capacitors are exploited to store the energy. Finally, the harvested energy is consumed by appropriate IoT devices for their corresponding applications [6].

However, several underlying challenges still hinder the realization of an efficient IoT harvesting system. For instance, the harvesting circuitry has a considerable impact on the hardware of an IoT object since the conventional IoT objects' designs are unable to handle the heavy fluctuations in an object's circuitry, primarily owing to the fact that the harvested energy delivered to an IoT object is predominantly reliant on the availability of energy within the environment and which occasionally could be either inferior or even superior to the power requirements of an object's circuitry. Similarly, intelligent software for IoT harvesting systems should be designed by the software developers which are capable of handling the energy's unavailability for a shorter duration of time to allow any task to resume and not restart from where it was left, thereby mitigating the data loss. Finally, both rechargeable batteries and super capacitors have some inherent limiting factors, and therefore, a highly efficacious, economical, miniaturized, and long-lived battery is still a challenge for the researchers.

B. Data-driven IoT

IoT offers the capability to connect and integrate both digital and physical entities. A fundamental challenge centers around managing IoT data especially when things are the majority of data producers and consumers. Given the intrinsic features of IoT data, topics such as storage, real-time data stream analytic, and event processing are all critical. Before diving into these four topics, we would first summarize these features.

Data generation in IoT has four main characteristics: i) Velocity--things produce data in different speed levels and some sensors can scan at a rate up to 1,000,000 sensing

elements per second3; ii) Scalability--IoT data are expected to be at an extremely large scale due to the ability of IoT sensors to continuously generate data together with the foreseeable excessively large number of things; iii) Dynamics--mobility is one characteristic of IoT things, leading to data generated in different locations under different environments at different times; and iv) Heterogeneity--many kinds of things have been and could be connected to the Internet and the data generated could be in different formats using different vocabularies.

The quality of the generated data usually faces some special challenges. Data could cause uncertainty and inconsistency as sensors and RFID tags would produce inaccurate readings and redundant readings, or even miss readings. Moreover, the data produced by assorted things can be interpreted in different ways, bringing challenges for proper interpretation of the produced data to different data consumers [7].

The nature of data produced by IoT calls for revisit of data storage techniques. Traditional datababase management systems (DBMSs) could be adopted for storing IoT data, but need to address the high processing and querying frequency. The development of large-scale, distributed storage systems is also raised to meet the exceptional demands of data storage in IoT and three factors need to be considered: consistency, availability and partition tolerance [8]. The storage issue in resource-constrained IoT scenarios also plays an important role due to the mobility and scalability of IoT data. Antelope4 is the first DBMS for resource-constrained sensor devices, which enables a class of sensor network systems where every sensor holds a database.

Linked Data5 is a method for publishing structured data and interlink such data to make it more useful. It builds upon standard Web technologies such as HTTP, RDF and URIs and extends these technologies to share information. Data from different sources can be connected and queried in the form of Linked Data. The concept of Linked Stream Data applies the Linked Data principles to streaming data, so that data streams can be published as part of the Web of Linked Data. SPARQL is a query language to query RDF database and has been extended to support streaming data querying [9]?[11]. Techniques in Linked Data perfectly fit the requirement of processing and querying IoT data at Web scale.

In IoT, complex event processing techniques lay part of the foundation of supporting computers to sense and react to events in the physical world. The focus of the complex event processing (CEP) model is on detecting occurrences of particular patterns of low level events indicating some higherlevel events, which are better understood by computers and humans. A semantic CEP system improves event processing quality by using event metadata in combination with ontologies and rules (i.e., knowledge bases). A knowledge base can be used to provide background knowledge about the events and other non-event resources [12]. With huge amount of external

3 4 5

domain knowledge available, information completeness and semantic matching are two research topics under semantic CEP. The process of reducing information incompleteness is called event enrichment which considers the challenges of enrichment source determination, information retrieval from the identified sources, and information fusion [13]. Semantic matching includes semantic selection which is evaluating pattern constraints based on the semantic equivalence of attribute meanings captured by the event ontology, and inexact selection which is selecting events while allowing a limited number of mismatches to detect relevant patterns [14].

Although we see prompt development of IoT techniques, many challenges and issues are worth to further explore from both academia and industry. Data quality and uncertainty remains a challenging problem due to the increasing data volume and heterogeneity. As in an IoT environment, the physical space and the virtual (data) space co-exist and interact. Data transferring, synchronizing and processing in the co-space demand novel techniques. Given various formats of external knowledge, semantic enrichment will not only deal with structured data, but also unstructured data at the same time. How to enrich semantic in IoT via hybrid formatted external information is also an interesting topic to explore. In contrast to utilizing external knowledge, discovering knowledge from the IoT data could contribute to the understanding of the IoT applications. Data mining techniques could be exploited for this purpose but in a distributed manner.

C. IoT Search

Searching and finding relevant objects from billions of things is one of the major challenges in the IoT era because the supporting technologies for searching things in IoT are very different from those used in searching Web documents due to tightly bounded contextual information (e.g., location) and no easily indexable properties of IoT objects. In addition, the state information of things is dynamic and rapidly changing.

By reusing techniques of the World Wide Web, the information and services of IoT objects can be provided on the Web [15]. This triggers the research of Web of Things (WoT) search engines (WoTSE), which is applying Web technologies to the Internet of Things to access information and services of physical objects. In WoT, each physical object possesses a digital counterpart that is commonly referred to as "Digital Twin". These digital twins are built according to Representational State Transfer (REST) architecture and accessed with HTTP protocol via RestFul API.

Research related to WoTSE begins from early 2000s and enjoys steady expansion ever since. It branches into different directions including object search, sensor search and functionality search [16]. In early projects, WoTSE are commonly used to locate physical objects, which are tagged with passive RFID tags or sensor nodes. Dyser [17] is one of the works that search physical entities based on their real-time states derived from their sensor readings. The work of CASSARAM [18] demonstrates the research effort on sensor search. It uses WoTSE for retrieving sensors based on their static meta-data

Fig. 2. Modular Architecture for Web of Things Search Engines [16].

and contexts, such as cost and reliability. Each form of WoTSE has its own characteristics, but all could follow the unified search architecture provided by our work in [16].

The modules in our architecture are organized into layers as illustrated in Figure 2. Two lower layers handle discovery activities while the two upper layers handle search activities. Storage modules for resource collections and indexes link two set of layers. The whole system is protected by security, privacy, and trust assessment measures, which are grouped into a vertical layer. To be more specific, the Discovery Layer serves as interfaces to the Web resources including sensor streams, representations, functionalities, websites and Web services. The Index Layer stores and indexes resources with its Collection Manager and Indexer modules. This layer also ranks the resources. The Search Layer carries out the query resolution process. The Query Processor module transforms raw user queries into the form processable by the system. The Query Dependent (Q.D) Ranker scores discovered query resources with respect to the user query and utilizes the recorded links between resources to find their corresponding result resources. The Ranking Aggregator module is responsible for combining different Q.D and Q.I ranking results into a final score for each resource. Finally, the Result Processor extracts and aggregates the information from matching resources and produces search results. The User Interface (UI) layer interfaces WoTSE with users. It provides Query Interface and Result Interface to receive queries and return search results, respectively.

The modular architecture provides a reference framework assessing the diverse implementation of the existing WoTSEs.

It assesses the support that each module receives from the existing works and how it is commonly implemented. The goal of WoTSE is building a search engine that could find anything available on the Web of Things. To achieve this goal, several challenges demand effective and efficient solutions. Crawling and indexing wide scale IoT data for the purpose of search are intrinsically problematic due to the dynamic and heterogeneous nature of IoT data. How to identify useful Web resources that are related to the things is also challenging as they could be in various format with different interpretation vocabularies as we discussed in the last section.

D. Security, Privacy, and Trust in IoT

The risk on data security and privacy exponentially increases with an unprecedented growth in the deployment of the smart IoT objects. One of the distinct challenges in the IoT infrastructures is the limited computation power and minimal resources of most of the IoT devices [19]. These limited resources preclude the state-of-the-art cryptographic techniques that are indispensable for securing IoT devices, thereby making them vulnerable to a diverse range of security attacks [20], such as the denial of service attacks and privacy attacks such as data exfiltration or leakage attacks.

Recently, there are numerous research proposals in the literature delineating on IoT security and privacy services such as [21]?[24]. Nevertheless, there are still open security gaps that require appropriate controls to mitigate them. The challenge is that the currently proposed systems do not provide a complete security solution that tackles all IoT security and privacy requirements. For instance, most of the proposed methodologies target one or two security requirements, e.g., confidentiality and authentication [25]. An efficient and reliable IoT data sharing requires an all-inclusive security solution for securing the data while limiting the interference that might occur if integrating several independent techniques to provide the required services. To the best of our knowledge, none of the proposed research methodologies or industry systems contribute a security attack free solution that provides conditional anonymous authentication and fine-grained access control techniques to be used by the resource-constrained IoT devices and infrastructures.

There are a number of challenges confronting the security of IoT infrastructure, including but not limited to, scalable security, denial of use of service or upload of data, and interoperability. Scalability is one of the indispensable requirement in the IoT infrastructures. Such a requirement can be met by delegating the expensive cryptographic computations in a secured manner to a cloudlet, edge, or cloud [26], [27]. There is thus a dire need for investigating intelligent ways to use edge computing with IoT and the cloud to address the current security challenges of IoT systems. Moreover, IoT security research studies should consider using cryptographic methodologies with limited communication overhead, such as constant size Attribute Base Encryption techniques [28], [29].

Non-repudiation is another essential requirement for IoT infrastructure, specifically for systems that include users'

interaction. Non-repudiation should be imposed to prevent users from denying either the use of the service or previous data upload. Unfortunately, non-repudiation is generally not considered in most of the current implementations owing to privacy concerns [30]. Several methodologies can maintain both users' and devices' privacy while implementing nonrepudiations such as conditional anonymity. Group signature [31], [32] is one of the techniques that can provide conditional anonymity. However, such techniques require extensive research to concur with the limited resources challenge within IoT infrastructures. Besides, interoperability is a vital IoT infrastructure requirement due to the heterogeneous nature of IoT devices. Considerable efforts and collaborations from governmental and non-governmental entities are required to create IoT interoperability standards and backward compatibility. These standards should also be integrated with privacy controls to guarantee the preservation of users' privacy.

Trust is also an indispensable issue in the IoT environment since the majority of the existing security mechanisms do not cater for the subjective belief among the heterogeneous IoT objects especially in the presence of internal malicious adversaries that intent to disrupt the reliability of a network [33]. Chaker et al. [34] delineates on trust as the degree of a subjective belief of an entity (trustor) over the other (trustee) in a specified context. IoT interplays between the paradigms of security and trust, i.e., if we regard security mechanisms in terms of barriers, locks, and accesses, then trust is a worry of when, where, and why to put these barriers, locks, and accesses in an IoT ecosystem to deal with the degree of collaboration and integration between the IoT objects [35]. Over the past decade, trust-based security mechanisms have emerged for enhancing the overall security of IoT, wherein trust and reputation models have been utilized to improve the collaboration and for selecting the trustworthy service provider based on quality-of-service (QoS), especially in service oriented architecture-based IoT [36], [37].

The importance of trust management has been recently investigated across numerous sort of networks, i.e., mobile ad hoc networks [38], peer-to-peer networks [39], social networks [40], and as-of-late for vehicular ad hoc networks within the context of the promising paradigm of Internet-of-Vehicles [41]. Evaluating trust becomes indispensable in the case of a highly dynamic and distributed network since pervasive infrastructure cannot be guaranteed at all the times in such scenarios which is imperative for public-key based cryptographic techniques. Nevertheless, computing trust has its own inherent challenges, i.e., selection of dynamic trust attributes in accordance with a given application's context, assigning of optimal weights to such attributes for trust aggregation purposes, opting between the event-driven, time-driven, or hybrid approaches for trust updates, and selecting an appropriate trustworthiness threshold for segregating between malicious and non-malicious nodes.

In essence, security, privacy, and trust go hand-in-hand for designing a resilient IoT network that could meet the stringent application requirements in realizing the formulation of highly secured digitized societies.

E. Service Computing and IoT

F. Social IoT

Initiated around the similar time as the Internet of Things, service computing (or service-oriented computing) has been established as an important paradigm to change the way of design, delivery, and consumption of software applications [42], [43]. Service computing relies on service-oriented architecture (SOA) and aims to organize software applications and infrastructures into a set of interacting services, which are then used as fundamental elements to support low-cost and efficient development of distributed applications.

Technologies on service computing (e.g., RESTful services and service composition methods) can help address several fundamental challenges presented by IoT including communication and management of IoT objects. However, marrying service computing and IoT presents challenges due to their technical constraints and unique characteristics [43]. On the one hand, IoT objects may be resource-constrained and the traditional service computing standards and techniques (e.g., SOAP, WSDL, BPEL) might be too heavy to be applicable in IoT. On the other hand, existing service composition models cannot be directly used for IoT interoperation, due to their architectural differences. More specifically, traditional service composition models are mostly single-typed and singlelayered (i.e., services), while IoT components are heterogeneous, multi-layered that include not only services, but also IoT devices and other components.

One important research direction centers on IoT services discovery, aiming to be able to find the right IoT services at the right time and the right location. There are two possible techniques. The first technique is semantic annotation for IoT service descriptions and their associated sensory data. Some typical efforts in this direction include the OpenIoT project6, which exploits a semantic sensor network (SSN) ontology from W3C, and the Hydra project7, which adopts OWL (an ontology for Semantic Web) and SAWSDL (a semantic annotation of WSDL). However, it is challenging to reach an agreement on a single ontological standard for describing IoT services, given the diversity and rapid IoT technological advances. The second technical direction is to use the textual descriptions associated with IoT devices to locate IoT services. Some typical efforts in this direction include MAX [44] and Microsearch [45]. One research challenge in this direction is the natural order ranking of IoT contents. Natural order ranking sorts contents by their intrinsic characteristics, rather than their relevance to a given query, thereby being able to deliver the most relevant results. One well-known example of natural order ranking is PageRank [46], which orders Web pages based on their importance via link analysis. Given the size of IoT (50 to 100 times bigger than the current Internet), one promising direction is to develop a new natural order ranking mechanism for the IoT contents in order to provide an effective and efficient IoT service discovery [2].

6openiot.eu 7hydramiddleware.eu

Recently, there have been quite a number of independent research activities to bring the next evolutionary step of the IoT paradigm by moving from smart objects to socially aware objects. This refers to creating a new generation of IoT objects that manifests themselves and have the capability to socialize with the surrounding peers mimicking human beings for the sake of, but not limited to, discovering new services, exchanging experience, and benefiting from each other capabilities. This new paradigm is referred to as the Social Internet of Things (SIoT), which is a new perspective that allows objects to establish their own social networks and navigate through the social network structure of the friend objects, allowing discovering other objects and their services.

Unlike the current process in IoT where search engines are employed to find services in a centralized way, SIoT can foster resource availability and make services discovery more easily in a distributed manner [47]?[49]. This paradigm also aims to provide reliable and trustworthy networking solutions by utilizing the social network structure. Based on the social structure established among IoT objects, objects can inquire local neighbourhood for other objects to assess the reputation of these objects and establish a level of trustworthiness. Additionally, SIoT enables objects to start new acquaintance where they can exchange information and experience.

SIoT is not a spur of the moment. There were earlier attempts to involve devices in the social loop. Back to 2001, Holmquist et al. [50] established temporary relationships between wireless sensors. In the work of [51], the authors discussed the idea of how objects can blog. Moreover, kranz et al. [52] enabled objects to share content using a social network framework Twitter. Guinard et al. [53] utilized the human social network as a framework for owners to share the services of these devices with their friends.

Previous attempts differ from the intended perspective of the current vision of SIoT. The current perspective refers to a new generation of IoT objects that have capability to form their own social network of friends without relying on the online human social networks. Several research activities have been conducted to realize this paradigm. In [47], Atzori et al. introduced this new paradigm and discussed the idea of integrating social networks concepts into the Internet of Things (IoT) for the purpose of addressing the related issues of service discovery and composition. They proposed a conceptual platform on how to enable IoT objects to create relationships among each other. They also identified policies of how to establish relationships between objects and how to manage these relations. Girau et al. [54] implemented an experimental SIoT platform. They evaluated the current implementations of IoT platforms and pointed out the major characteristics that can be reused in this experimental SIoT platform. It includes several functionalities that can enable the smart objects to register into the platform as a first step. Then, the system manages the creation of the new relationships. Using this system, smart objects are capable to create groups of members with similar characteristics.

 ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download